Trust for autonomous intelligence

MOSS gives each agent a cryptographic identity and signs every output. If four agents touched a payload, you can trace exactly which one produced what — and verify nothing was tampered with in between.

We use ML-DSA-44 (FIPS 204) because RSA, ECDSA, and Ed25519 won't survive quantum computers, and audit logs need to stay valid for decades. Existing tools like GPG weren't built for agents, and JWTs are auth tokens, not signatures.

from moss import Subject

agent = Subject.create("moss:app:agent-1")
envelope = agent.sign({"action": "approve"})

# Verify origin
assert agent.verify(envelope)

$ moss keygen --out agent.key
$ moss sign --in data.json --key agent.key --out data.sig
$ moss verify --in data.json --sig data.sig

Integrations for CrewAI, AutoGen, LangGraph, and LangChain.

A stateless, one-shot authentication proof built from two-mode squeezed entanglement and homodyne measurements.

Two-mode squeezed states (TMSS) and homodyne detection are mature, high-rate, room-temperature tools in quantum optics. SEEP explores whether they can provide authentication where there's nothing to extract — each proof consumes the shared quantum state.